under the hood
Models we use
takez runs entirely on Anthropic’s Claude models — but not one-size-fits-all. Each part of a take gets the right brain for the job.
Claude Sonnet 4.6
powers your Corner
The conversation model. Fast and sharp, it's what you talk to when you build your case — turning plain language into your facts, your reasoning, your hypotheticals.
Claude Haiku 4.5
powers the swarm
The lightweight workhorse. When the Consensus fans out up to 500 agents to research, cross-examine, and fact-check every claim, this is what makes that swarm affordable enough to be silly with.
Claude Opus 4.8
writes the Verdict
The heavyweight. Anthropic's most capable Opus-tier model weighs the locked cases, the swarm's findings, and the dissent — then writes the ruling and its confidence score.
Why it matters: looping hundreds of small agents that check each other’s work beats one big model answering once. The swarm does the legwork, the heavyweight makes the call.